[Bf-committers] Automated test harness

bf-committers@blender.org bf-committers@blender.org
Fri, 28 May 2004 00:23:30 EDT

Hello folks -
I mailed this to Ton:
> I'd like to work on stuff to help automate blender testing.  One  thing  
>  was discussed in one of the weekly Sunday meetings a couple  of
>   weeks ago was a way to save a baseline file from a render, and 
> then be able to re-render and do a comparison. 

>  - Do you think this would help the most?  Or would an effort on
>  automating GUI testing help find more bugs/regressions? -  If two
>  renderings are not the same, we will need to report  the difference.
>   Do you think a visual comparison is best (e.g., a   "NOT-AND" type
>  of operation between the two bitmaps to show where  the differences
>  show up) or some other sort of comparison?
> - Willian  said that you have some kind of regression suite.
>  How do you run those tests?  Could I get some of those  files
>  if I decide to go off and work on the  render-comparison?
and got this response:
> Here's the test suite:
_>  http://download.blender.org/demo/test/_ 
> I thought of using the 'swap page' feature in Blender (Jkey), to  use  
> the default available 2nd buffer for comparision. However, I  also think  
> the testing moethods shouldn't overdo for now. Looking  at the past, the  
> most critical slips are in  tools/hotkeys/commands. Having a clear  
> scheduled public release  program might help in attracting more people  
> to use the trial  versions as well.
Is there a consensus that time is better spent on setting up a GUI test  
harness rather than a render test harness?
Thanks much,
Chris Keith